Abstract

Condition-based maintenance (CBM) is an effective maintenance approach to prioritize and optimize maintenance resources based on condition monitoring information. A well established and effective CBM program can eliminate unnecessary maintenance actions, lower maintenance costs, reduce system downtime and minimize unexpected catastrophic failures. Most existing work reported in the literature only focuses on determining the optimal CBM policy for single units. Replacement and other maintenance decisions are made independently for each component, based on the component’s age, condition monitoring data and the CBM policy.
In this thesis, a CBM optimization method is proposed for multi-component systems, where economic dependency exists among the components subject to condition monitoring. The proposed multi-component systems CBM policy is based on a method using artificial neural network (ANN) for remaining useful life (RUL) prediction which is proposed by Tian et al. (2009). Deterioration of a multi-component system is represented by a conditional failure probability value, which is calculated based on the predicted failure time distributions of components. The proposed CBM policy is defined by a two-level failure probability threshold. A simulation method is developed to obtain the optimal threshold values in order to minimize the long-term maintenance cost.
We conduct a case study using real-world vibration monitoring data to validate the proposed CBM approach. These data are collected from bearings on a group of Gould pumps at a Canadian Kraft pulp mill company and help to demonstrate the effectiveness of the proposed CBM approach for multi-component systems. The proposed CBM approach is also demonstrated using simulated degradation data for
multi-component systems. The proposed maintenance policy can fulfill the requirements of a real plant environment where multiple components are under condition monitoring. By using the proposed CBM policy, maintenance managers can easily and quickly adjust the maintenance schedule according to the working condition of the system.